How to Develop Leverage for Safety Improvement

Brooks Carder, Ph.D.
Business Intelligence
a division of AdGap

The World Today
It seems today that there is a new fad in management theory coming along every day, pushed by some guru or another. The theories and recommendations are often contradictory. Certainly if you read Fortune or Business Week it is easy to become confused. Some things I believe, however, are virtually certain. Competition will increase. We will have to do more with less. We will have to be more efficient in our use of capital and human resources. In short, we have to "get better or get gone." This applies to safety as well as to every other aspect of business. It may even apply to government some day, but don't hold your breath.

In safety, we need to reduce incidents and their financial and human cost while spending less and not interfering with production and quality. I am going to tell you about an approach that will meet these demands. It is based on our research and consulting work over the past six years.

As I see it, there are three ways that you can approach the problem of safety improvement. Two hold promise and one is doomed:

1. Do what you are doing now but do it more vigorously.
2. Observe and attempt to modify worker behavior.
3. Use the Plan-Do-Study-Act cycle of Total Quality Management.

The first is the general approach of most companies who attempt to "tweak" their program for improved awareness, etc. If your system is stable, as defined by an accident record that demonstrates a process "in control," this approach will not work. Deming has emphasized that a stable system will not change without intervention from outside.

Another problem with the first approach is that it causes safety to compete with other priorities. If you increase emphasis on safety, you must decrease emphasis elsewhere. Sooner or later, the tables will turn.

The second approach has promise. It is currently very popular. The data I have seen show improvement in a number of companies. However, I see three obvious limitations: 1) It is very expensive. 2) It is difficult to implement in environments where workers work alone or in very small groups spread out over a wide area. 3) It has limited effectiveness on problems of process safety. Process safety relates to processes like chemical manufacturing and petroleum refining that can "run away," causing major destruction and loss of life. Process safety is very hard to relate to observable behavior.

I am here to talk about the third approach, which I strongly favor.

The PDSA cycle, sometimes called the Shewhart cycle or the Deming cycle, is the fundamental method of TQM for the improvement of a "stable" system. It begins with a process of developing a "Plan." This process involves collecting and analyzing data about the system. The sources of data should include the opinions of people who work in the system. There should also be other sources of system evaluation. Based on these data, an action plan is established.

In the "Do" phase, the plan is implemented, on a small scale if possible.

In the "Study" phase, the results of the intervention are studied. Again, a variety of data sources should be available. As with the planning phase, measurement should be carried out with adequate knowledge of statistics and scientific method. Based on the results, the plan may be revised.

In the "Act" phase, the effective aspects of the plan are implemented on a wider scale. The cycle repeats again and again to generate continuous improvement.

In this paper, I will describe some of the theory and method behind this approach by answering three questions:

1. What does TQM have to do with safety?
2. What is the "safety system" and how do we understand that system in order to develop a plan?
3. What results have been obtained by organizations that have taken this approach?

Safety and Quality are One in the Same
Since 1989, our group has been involved in the application of TQM principles to the problem of improving safety performance.

Your company produces something: a product that you sell, a service that you sell, or both.

It also produces accidents. It probably produces them reliably. You do not know when they will happen, but with a large unit over a period of time, incidents will occur and the frequency will probably remain within statistical limits.

If we plot the number of accidents each month on a chart, and calculate statistical "control limits," we usually find a stable process. This means that the variation in the process can be ascribed to random, statistical fluctuation, not to changes in the system. If a system produces an average of four incidents in a month, the laws of variation tell us that the system could produce as many as ten or as few as zero accidents without changing.

Accidents in and of themselves are evidence that you do not have control of the process that you are operating. They are an additional but important source of information about defects in your operation.

It is unlikely that a unit with an unusually high frequency of accidents will be maximally effective in either productivity or quality. The accidents are evidence of inadequate predictability and control in the organization. The defects that produce accidents will also impair productivity and quality.

Let us look at one of AdGap's current clients in the manufacturing sector. We studied thirteen plants within the organization. Seven were selected because they had low accident rates and apparently good safety cultures. Six were chosen because they had frequent incidents and apparently weak safety cultures.

In the course of our study, we found that a staff group within the manufacturing organization had studied these same thirteen plants, (along with all the company's other plants), to determine the effectiveness of each plant's management system for producing quality and productivity. We asked for that group's ratings of our thirteen plants. The scatter plot in Figure 2 demonstrates the correlation between the two measures. Plants with weak management systems had higher accident rates.

Productivity is not achieved through strong motivation and cutting corners. It is achieved by a well-trained and well-led workforce performing according to well-designed plans and procedures.

Reactive Approaches and the TQM Method
In companies that employ strategy #1, (Do what you are doing more vigorously), accidents give rise to reactions designed to prevent their recurrence (and usually to fix blame). While well-intentioned, these actions guarantee the future occurrence of similar, if not identical incidents. If you disagree, and believe that these actions truly are effective, I would ask you which types of incident you have eliminated. Which types have you substantially reduced? If you have a lot of success, you are way ahead of most organizations.

The reason that these actions are not effective in the long-term is that they are not based on a true understanding of the causes. Human error is not a useful finding of cause. Why is it that the humans in some organizations make more errors than the humans in others? Why do Ford workers make fewer errors than GM workers? Why do George Siefert's players make fewer mistakes than Jerry Glanville's?

For many years, American car makers used elaborate inspection systems to discover and eliminate defects. The humans who made these defects were often disciplined, and fixes were designed to eliminate the defects forever. The fixes failed and the defects remained. The inspection and control systems added cost and no longer produced value.

Toyota was among the first companies to benefit from the application of TQM principles as taught by Dr. W. Edwards Deming, Joseph Juran and others. TQM assumes that defects come from the system, not from the people. While the people are part of the system, most of them do not control it. The system is largely controlled by management. Improvement of quality can only be achieved through a process of understanding and improving the system.

Figure 3 shows the results of the application of TQM principles to the manufacture of automobiles. The first set of data was compiled by an MIT group that conducted a massive study of auto manufacturing in the late 80's and early 90's. It is obvious that if GM didn't change, their goose would be cooked.

Was this because, as some would say, the Japanese are smarter, have a culture that is more conducive to quality, or something like that? No! Ford saw the handwriting on the wall in 1981, and they adopted TQM. In fact, when the MIT study was done, the best plant they found was not a Japanese plant, but a Ford plant in Hermisillo, Mexico. It isn't the people. It is the system.

Figure 4 compares the productivity of the three U.S. automakers, before and after the adoption of TQM. Ford was the leader in this, while GM trailed behind.

Figures 5 and 6 show the impact on the bottom line, and it is getting even worse for GM. In the third quarter of 1994, GM lost over $800 million in its North American automotive operation, in spite of the fact that market conditions were excellent! Ford made over $1 billion in the same period.

At this point, you may want to ask what the quality of automobiles has to do with safety, or anything else in the communications business. It has everything to do with it! First, effective management systems generate superior performance, and that performance includes better safety performance. With a superior management system, people are doing the right thing in the right way, at the right time, with the least effort and the greatest impact.

Second, the basic principles of TQM apply to any organization: a manufacturing plant, a football team, or a communications company.

Jerry Glanville motivated his players with inspiring speeches. He encouraged them to try hard, and they usually did...in the first half. Bill Walsh motivated his players by giving them a superior system and teaching them to execute. Nothing motivates like success.

TQM has two basic assumptions:

1. People will usually do their best to do what they believe is expected of them. The job of leadership is to clearly communicate what is expected (with actions as well as words) and to make certain that the people have the tools needed to do the job. I do not rule out discipline, because ultimately I believe that some is necessary. However, if a great deal of disciplinary action is necessary, then I would want to know what is really being communicated regarding expectations. For example, if bending the rules for success is considered heroic until someone gets hurt, the message is, "It's OK to bend the rules, just don't get hurt."

   It is gratuitous and perhaps insulting for management to tell employees not to get hurt or not to make a mistake. Only psychopathic people want to get hurt. You should be able to eliminate these people quickly.

2. Defects (accidents, errors, etc.) come from the entire system of people, policies, procedures and practices. Reducing defects requires an understanding of the root causes of those defects in the system. In practice, this usually requires study and the application of formal (scientific) knowledge. Very often, the obvious solution, the one that is not based on adequate study, will just make things worse.

All of this is well and good, and many people say that they grasp the principle, only to start digging harder but calling it TQM. I have seen a lot of failures, and I have seen a lot of cynical employees who understand that management is announcing change, but that management is not changing.

The failure lies in a failure to understand the theory of systems.

The Theory of Systems
A system is a set of interrelated components which function together to create an output. The output is not just the arithmetic sum of the outputs of the parts, but is a result of the interaction of the parts as well. If interaction doesn't matter, you don't have a system. In a baseball team, the interaction of the players is relatively less important. In basketball, it is more important, and in football, it is most important. In a business it is very important. Deming estimated that in an organization like a business, the system would account for a least 85% of the variation in performance, while the individual contributions could account for no more than 15%.

Safety, or accidents if you wish, are the output of a system of extraordinary complexity. The system is far too complex to understand in detail. It is far too complex to allow the prediction of the specific time and place of an accident. However, through an understanding of systems theory, and an examination of an organization, we can identify patterns that allow us to gain some understanding of the dynamics of the system. Based on this understanding, we can develop actions that will change the output of the system in the manner we wish. If we fail to understand the system, our actions will often make things worse.

The Principle of Leverage
In order too improve the system, we need to understand it. If we understand it, we can identify the leverage points - the points where our inputs will achieve maximum results. If you don't find the leverage points, then at best you will have to spend a great deal of effort to accomplish change. More likely, you will achieve no results at all. Systems have a way of pushing back at us as hard as we push on them. The more effort we put into interdicting the drug supply, the more profitable the business and the more incentive there is to engage in it. Perhaps there is a sufficient level of force to shut it off, but we are definitely not working on a leverage point. (Mainland China and Japan both dealt successfully with drug abuse epidemics by suspending any due process and dealing severely with the users.) Legalizing drugs might reduce profits and incentives to sell. But it would also probably increase use. In the end, profits might stay just as high.

Raising taxes to increase revenue may lead to people finding loopholes and/or moving away, so that revenue does not increase.

We recently conducted an assessment of the safety system in a chemical plant with a fairly good safety record. In looking for instances where procedures were not followed, we discovered that the mixing recipes would be changed when product had to be rushed to the customer. Instead of adding each ingredient in a timed sequence, the workers would dump in all of the ingredients at once and cook the batch. The workers claimed this did not cause any problems.

In fact, product quality was an issue at the plant. Management was unaware of the violations that we had discovered.

The obvious solution was to round up the employees and insist that they follow procedures. We cautioned management against this. Instead, we looked at the entire system and culture.

The success of the company had been built on its attention to "taking care" of customers. We did not want to discourage this, or to fight against it. Therefore, we looked into the source of the rush situations. A major cause was a slowness in paperwork getting through the administrative system. Orders were taking as long as two weeks to reach the production floor, creating rush situations. We suggested working on this, as a leverage point.

Two years later, the accident rate of this plant has been reduced by 50%. The major effort was the involvement of quality teams to improve processes in the plant, like order entry. This improvement is not the result of a safety awareness campaign. It was not costly. It led to improvements in quality and productivity as well as safety.

System and Culture
When we talk about the "safety system," we are talking about the entire set of policies, procedures and practices that influence safety. Consider the diagram in Figure 7 which represents a portion of the safety system. It is obvious that there are many remote influences which are very important.

While the "architecture" is extremely complex, it is relatively concrete and easy to change, at least in principle. We can quickly change policies and procedures, at least in theory. Changing practices takes a bit longer. The greatest difficulty arises, however, when we consider culture. Operating over time, a system creates a culture. The culture, in a simplified view, is the set of attitudes, beliefs and expectations that drive behavior. These are not easily changed. Just because leadership says something has changed doesn't mean that workers will believe that it has changed (or if it has, that the change will last). As the size of the federal bureaucracy (or any bureaucracy) grows, it carries with it a culture that is quite difficult to change. Leadership can change the policies and the recommended procedures. Practices are much more difficult to change, especially if the culture runs contrary to the desired change.

What I am saying is that it is not difficult for leadership to spell out where it would like to go. The problem is getting there. We can try to get there with tremendous control and coercion. This rarely works with people. The alternative is to design leveraged actions. This requires an understanding of the dynamics of the particular system and culture that we want to change.

Measurement of the System and Developing Leveraged Actions
Our group has successfully employed two methods for the measurement of a system, including culture:

1. Written surveys. The Minnesota Perception Survey and its variants are written surveys developed by a group of railroad safety managers and by scientists at the Aberdeen Proving Ground in the early 1980's. This process has been used successfully by scores of companies to improve safety performance.
2. In-person interviews. Management systems can be studied by interviewing employees in order to understand the dynamics of the management system. This process has been used successfully by scores of companies to improve safety performance.

Each of these approaches has substantial validity. Let me explain "validity" further. If a measure is valid, it means that you are measuring what you wish to measure. I consider three levels of validity:

1. Face Validity. Does the measurement "make sense." For example, a question like "Did you receive adequate safety training?" has face validity. A question like "Do you like apple pie?" has no face validity in measuring safety. However, it could have the next two kinds of validity.

2. Predictive Validity. This is the ability of your measure to predict or correlate with other measures of the same thing. The prototype here is the IQ test, which is an attempt to predict something like "ability to learn." In fact, IQ tests do correlate with school performance. However, the correlation is quite weak, with a correlation coefficient on the order of 0.2 to 0.3. This means that an IQ score accounts for between 4% and 10% of the variation in school performance. The rest of the variation comes from something else.

   We would want our measures of the safety system and culture to correlate with some measure of loss. In industry, we use the recordable accident rate. If we have two sites with relatively equal risk, the one with the better score on our measuring instrument should have fewer accidents.

3. Theoretical Validity. Here I ask if the measure provides us with sufficient understanding to intervene successfully. The IQ test has little or no theoretical validity. It does not tell us anything about the source of problems that might be detected, or what to do to improve. All three forms of measurement described above, (written surveys, in-person interviews and direct observation of behavior), can have theoretical validity. The proper use of each of these measures has led to improvement of safety performance.

Let me add one important thing here. For the perception survey and the assessment of the management system, the most valid opinions are the opinions of hourly employees, not managers. Statistical analysis of managers' perception survey scores shows no correlation with performance. When Dow first developed their management system assessment, it was scored by managers and it was not effective. When the survey was put into the hands of employees, the resulting evaluations began to assist in the improvement of safety performance.

The Perception Survey
I want to concentrate for a while on the perception survey, to show you how we use it and what it can tell us. As I mentioned earlier, the survey was developed by a group of railroad safety personnel and scientists at the Aberdeen Proving Ground. It has subsequently been applied to over 100,000 employees in over 50 companies. We have studied over 10,000 employees in over 20 companies.

So that you can see what the survey is about, I am listing a few of the questions below:

• Does management insist on proper medical attention for injured employees?
• Do supervisors pay adequate attention to safety?
• Is safe work behavior recognized by supervisors?
• Do employees participate in setting goals for safety?
• Does the company seek prompt correction of hazards found in inspections?
• Is safe work behavior recognized by your company?
• Have your company's efforts encouraged you to work more safely?
• Are maintenance programs at a level which help prevent accidents?

The survey that we use consists of 32 validated questions, derived from the original 74 in the Minnesota Perception Survey. I have validated the survey in three different business environments; a chemical manufacturing company, an equipment manufacturing company and a hazardous waste disposal company. In all three environments, units with higher accident rates have lower perception survey scores. The correlations that we observe range from 0.6 to 0.8, meaning that we are accounting for between 36% and 64% of the variance. (Remember, IQ tests account for less than 10% of performance measures.)

Of course, it could be that if a lot of accidents are happening, this causes lower perception survey scores. In other words, correlation between survey scores and accident rates is not proof of a causal relationship.

However, we have demonstrated that interventions directed at correcting deficiencies identified by the survey not only change the survey scores, they change the accident rates as well. This is evidence that the perception survey is measuring underlying factors in the causation of accidents. I will discuss some of our leverage interventions later.

What does the Perception Survey Measure?
There is a danger in using the perception survey. With 32 or 74 questions, you could accumulate a very long list of things to fix. It has been our hypothesis that the results of these individual questions are themselves symptoms of some more fundamental issues. Our research has been directed at identification of these "fundamental" factors measured by this survey. In order to develop leveraged action, we must work with fundamental issues, not symptoms.

For over two years we have been working with Rhone-Poulenc, a large French-owned chemical company with over 50 plants in the US. In May and June of 1994, we administered the Minnesota Perception Survey to over 6,000 U.S. employees. In addition to validation studies, we performed a factor analysis on the data. This is a statistical analysis designed to identify sets of questions that correlate with each other. The idea is that these sets of questions, or factors, are measuring the same thing. Having identified sets of questions or factors through statistical analysis, you then look at the content of the questions to see what each factor might be measuring.

In our analysis we identified the following four major factors:

1. Leadership. This is management's demonstrated commitment to safety. It includes things like recognition, promptly correcting hazards that employees identify, and having supervisors that "pay attention" to safety.
2. Education and Knowledge. This covers a variety of aspects of training, including job training and safety training. It also covers knowledge, such as an understanding of the hazards of the work.
3. Quality of the Safety Supervisory Process. This relates to how the company ensures that things are done the way they are supposed to be done. It covers issues like how regulations and procedures are enforced, and how standards are communicated.
4. Employee Involvement and Commitment. This measures a variety of involvement activities, such as employee participation in inspections. It also measures commitment, such as co-workers' support for the safety program.

These factors are not unique to Rhone-Poulenc. We subsequently performed the same analysis on survey results from a unit of Eastman Kodak with which we were working. The same four factors emerged.

The Perception Survey and Leveraged Action
The utility of this form of analysis emerges when we begin to use it to identify fundamental weaknesses of a particular unit. We have extensive normative data of the questions, so that we can determine how an organization compares with its peers on each of the factors. Areas of weakness are the target for leveraged intervention.

Two examples of leveraged action:
Through assessments of a large pipeline company, we identified a lack of leadership support for safety. The problem was not in leadership's attitude, but in its active demonstration of support. One of the ways that leadership was not demonstrating support was that there was no recognition of safety contributions. We are not referring to recognition for not having an accident. That can be a result of luck. Recognition should come from a direct observation of safe work or of efforts to improve safety.

We helped this company introduce a program in which supervisors were trained to identify and give positive recognition for contributions to safety.

Accident rates were recorded in control charts over a period of several years. Following the intervention, we observed an improvement in the leadership scores on the perception survey and a 50% reduction in the accident rate, which has persisted for over three years.

Here we have an example of a leveraged solution. We didn't need to chase accidents or symptoms. By examining the system, we developed a solution that was easy to implement, and that had powerful results.

In a different organization, with several thousand employees and more than a score of plants, we developed an analysis of strengths and weaknesses for each site, and for the organization as a whole. (Note that a plant site may be strong in leadership, in spite of the fact that the overall organization is relatively weak in that area.)

Based on our findings, we held meetings with representatives of each plant site and with the top management of the organization. The objective of the meetings was to provide them with the data for their organizational unit and to teach them about the concept of using these findings to develop leveraged actions. Each site was responsible for formulating an action plan based on the data.

The organization is still in the process of developing and implementing actions. However, as the diagram in Figure 8 shows, we have already seen improvement in the safety record of the organization. This diagram shows a control chart running from the last half of 1990 through January of 1995. There was a process shift around the end of 1991, with the mean going from just above 5.0 to just above 3.0. The process remains in control for about two and one-half years, until about June of 1994. Following June, we have seven consecutive months below the mean of 3.02. This indicated another process shift, with the new mean at 2.30. The obvious "special cause" is the company-wide administration of the perception survey and the related activities that followed. This process began in April of 1994, and is still con-tinuing. While this seems to be a very rapid impact, we have seen similar rapid improvements in other or-ganizations.

All of this work is highly leveraged. It has been very inexpensive. It has not interfered with production or quality. It has been very well received at all levels of the organization. We are aggressively continuing the process at both the plant and corporate levels to generate further improvement.

Conclusions

First, I admit to trying to condense a great deal of thought and study into a brief presentation. I have tried to accomplish the following:

• To explain the principle of leverage.
• To show how leverage can be developed through the study of a system.
• I have argued that leveraged solutions are always the most efficient. In many cases, no progress can be made without leverage.
• I have discussed some methods for study of the system: written surveys and in-person interviews.

The principles that I have discussed can improve any or all of the outputs of the system. Improvements in safety, produced through leveraged solutions rather than brute force, will lead to improvements in overall performance. The solutions that derive from our methods will help you to get the right person to do the right thing at the right time. That is the essence of leadership.

For Further Reading

Ackoff, Russell, Creating the Corporate Future. Wiley, New York, 1981.

Deming, W. Edwards, Out of the Crisis. MIT Center for Advanced Engineering, Cambridge, 1986.

Deming, W. Edwards, The New Economics for Industry, Government, Education. MIT Center for Advanced Engineering, Cambridge, 1993.

Imai, Masaki, Kaizen. Random House, New York, 1986.

Ingrassia P. and J. B. White, Comeback: The Fall and Rise of the American Automobile Industry. Simon and Schuster, New York, 1994.

Kaufman, Draper L., Systems One: An Introduction to Systems Thinking. Pegasus Communications, Cambridge, 1980.

Senge, Peter, The Fifth Discipline. Doubleday, New York, 1990.

Senge, Peter, et. al., The Fifth Discipline Fieldbook, Currency Doubleday, New York, 1994.

Waldrop, Mitchell, Complexity: The Emerging Science at the Edge of Order and Chaos. Simon and Schuster, New York, 1992.

Walton, Mary, The Deming Management Method. Perigee, New York, 1986.

Wheatley, Margaret, Leadership and the New Science. Berrett-Koehler, San Francisco, 1992.

Whitney, John 0., The Trust Factor, McGraw-Hill, New York, 1993.

Womack J. P. et. al., The Machine That Changed the World. New York, Rawson Associates. 1990.